TencentOS Server Full-Stack Ops Diagnostics Expert

Problem

TencentOS Server incidents often span disk, network, performance, memory, system, security, and CVE issues. A single check script may miss cross-layer symptoms, while loading every operations document at once can quickly consume context. This skill treats diagnostics as a staged workflow: match the reported symptom to a capability index first, then load only the relevant reference documents for targeted troubleshooting.

How it works

After triggering, it loads guides/workflow.md, guides/capability-index.md, guides/scenario-routing.md, and guides/doc-map.md in order. workflow.md covers dependency setup and diagnostic steps, capability-index.md maps the problem to modules, scenario-routing.md routes symptoms to the right module, and doc-map.md helps locate subdocuments. Once a module is matched, it loads only 1–3 related references/ documents instead of expanding all materials. Performance analysis prefers perf-prof, falling back to native perf commands when perf-prof is unavailable. If the index has no match, it can search keywords inside references/.

Boundaries and notes

The skill is scoped to TencentOS Server operations diagnostics, covering 24 capabilities across disk, network, performance, memory, system, security, and CVEs. Requests about web/mobile development, UI/frontend optimization, websites, CMS, SEO, business code, data modeling, application SQL design, AI model training, general data analysis except GPU fault diagnosis, programming teaching, or product requirements should be declined. Credential handling is strict: the assistant should not read ~/.ssh/id_*, repeat pasted passwords or private keys, or write credentials into scripts, logs, or commits. Users should run ssh themselves, while the assistant only generates an ssh user@host template. If plaintext credentials are pasted, flag the exposure and recommend rotating keys.